Effects of screening colonoscopy in preventing colorectal cancer: 
re-analysis of the NordICC trial

doi:10.21203/rs.3.rs-4048301/v1

Background. In the NordICC trial, the first randomized trial reporting on the effects of screening colonoscopy, reduction of colorectal cancer (CRC) incidence during 10 years of follow-up was estimated as 18% in intention-to-screen analysis, and 31% in adjusted per-protocol analysis. However, the „incident“ cases considered in these analyses included large proportions of CRCs that were already prevalent at recruitment and could not possibly have been prevented by the screening. We derived effect estimates for the reduction of truly incident, actually preventable cases from the published trial results.

Methods. Numbers of truly incident cases, which were directly observable only among screening attenders, were estimated for all groups of participants under the following plausible assumptions: (i) equal CRC prevalence and incidence in the invited and the usual-care group in the absence of screening effects (plausible due to the randomized design and the large sample size), and (ii) equal share of prevalent CRC cases among all CRC cases in screening non-attenders and attenders in the absence of screening effects (assumption needed for intention-to-screen analyses only). We calculated screening effects with respect to reduction of truly incident cases.

Results. With relative risk estimates (95% CI) of 0.72 (0.59-0.91) and 0.43 (0.26-0.78), offer and use of screening colonoscopy were estimated to reduce truly incident cases by 28% in intention-to-screen analysis and by 57% in adjusted per-protocol analysis.

Conclusion. The ability of screening colonoscopy to prevent incident CRC is much stronger than originally reported from the so far only RCT on long-term effects of screening colonoscopy.

Based on intriguing findings from observational studies [1,2], colonoscopy has since long been recommended for colorectal cancer (CRC) screening [3], long before evidence on its effectiveness in reducing CRC incidence and mortality was demonstrated by a randomized controlled trial. Such evidence has only very recently been provided by the Nordic-European Initiative on Colorectal Cancer (NordICC) trial [4]. In this randomized trial, the risk of CRC was lower among those invited to undergo screening colonoscopy than among those not invited to screening. However, reported CRC risk reduction was substantially smaller than anticipated, with risk ratios (95% confidence intervals) of 0.82 (0.70-0.93) and 0.69 (0.55-0.83) in intention-to-screen analysis and adjusted per-protocol analysis, respectively. These results have prompted major concerns and debate about the effectiveness and cost-effectiveness of screening colonoscopy [5,6], which had so far been considered as a particularly effective and cost-effective, if not cost-saving CRC preventive measure [7]. Also, the trial results seemed to be in conflict with the dramatic decline in CRC incidence that has been achieved in the screening-age population in countries with widespread use of screening colonoscopy, such as the US [8].

One explanation for this discrepancy is that the analysis of the NordICC trial included a large number of prevalent CRC cases, i.e. cases that were already present at recruitment. However, there is no way screening could have prevented such prevalent cases. The reason why prevalent cases could not easily be excluded in this trial is that prevalent cases were only known with certainty for participants who underwent screening, but not for non-attenders. We previously demonstrated that including prevalent cases strongly diluted the effect estimates of screening colonoscopy regarding prevention of truly incident, potentially preventable cases [9-12]. However, the extent of such dilution had remained rather uncertain. In a re-analysis of the NordICC trial, we aimed to derive, using well-defined, transparent and plausible assumptions, unbiased point and interval estimates of the reduction of truly incident, actually preventable CRCs by screening colonoscopy.

Study design and study population

Details on the design of the NordICC trial have been reported elsewhere [4]. Briefly, this trial was run in four Northern European countries, Poland, Norway, Sweden and the Netherlands, but data from the Netherlands were not included for legal reasons in the first report on long-term effect estimates, published in October 2022. The study population for that analysis, which was drawn from population registries, included 84,585 presumptively healthy men and women 55-64 years of age who were randomized in a 1:2 ratio to either receive an invitation for a single colonoscopy or to usual-care between 2009 and 2014. Out of 28,220 participants in the invited group, 11,843 (42.0%) followed the invitation. Primary endpoints were risk of total CRC incidence and death. Follow-up was performed by record linkage with cancer registries and cause-of-death registries. During a median follow-up of 10 years, 259 and 622 CRC cases were registered in the invited group and the usual-care group, respectively. Among the 259 CRC cases in the invited group, 102 were registered among attenders of screening colonoscopy, of which 62 were prevalent, screening-detected cases and 40 were truly incident cases (Table 1, upper part). The numbers of prevalent and incident cases could not be directly determined for non-attenders of screening and for participants in the usual-care group who did not receive an intervention.

Statistical analysis

The aim of this statistical analysis was to derive estimates of relative risk of truly incident CRC even though these cases could not be directly observed among non-attenders of screening and for participants in the usual-care group. Just excluding known prevalent CRC cases among screening attenders from the outcome measures would not be an option because selective exclusion of such cases for screening attenders would lead to overestimation of screening effects. Thus, it is necessary to also estimate the numbers of both prevalent and incident CRC cases among non-attenders of screening and participants in the usual-care group and then carry out an additional analysis after excluding prevalent cases from event numbers for all groups of participants.

The assumptions made in our analysis are listed in Table 2. The first assumption is that equal proportions of prevalent and incident cases in the invited group and the usual-care group would be expected if screening did not have any effect. This assumption should be ensured by the randomized study design and the very large number of study participants in each group. The second assumption, which is required for intention-to-screen analyses of reduction of incident cases only, expects equal shares of prevalent CRC cases among all CRC cases among non-attenders and attenders of screening in the absence of screening effects, which also appears plausible. Note that this assumption does not require equal CRC risks among attenders and non-attenders of screening. Furthermore, we conducted sensitivity analyses to evaluate the potential impact of even major violation of this assumption.

Step 1: Calculate the expected number of CRC cases prevented by screening:

From the published data, we first derived the total number of CRC cases that would have been expected in the invited group if screening did not have any preventive effect. This number was derived by multiplying the number of participants in the invited group with the apparent cumulative incidence in the usual-care-group: 28,220 x (622/56,365) = 311. As only 259 such cases were observed in the invited group this suggests that 311-259=52 CRC cases have been prevented by screening (Table 1, middle part).

Step 2: Calculate the number of incident cases that would have been expected in the screened group, if this group had not been screened:

The prevented cases would be expected to otherwise have occurred as incident cases among screening attenders, which would have resulted in 40+52=92 incident CRC cases. Together with 62 prevalent (screening-detected) CRC cases, 92+62=154 total CRC cases would have been expected in the screening attenders in the absence of screening effects (Table 1, middle part).

Step 3: Calculate the expected number of prevalent and incident cases in screening non-attenders and in the usual care group:

Assuming the same shares of prevalent and incident cancers among all cancers in screening non-attenders and attenders if screening had no effect, i.e. 62/154=0.40 and 92/154=0.60, respectively, the expected numbers of prevalent and incident CRC cases can be derived as 157×0.40=63 and 157×0.60=94 in screening-non-attenders and as 622×0.40=250 and 622×0.60=372 in the usual-care group (Table 1, middle part).

Step 4: For the screened group, use the actually observed values for prevalent and incident cases:

From the so-derived expected numbers of prevalent and incident CRC cases in the absence of screening effects we derived the corresponding numbers with screening effects by replacing the respective numbers in screening attenders by the observed, lower numbers. The numbers of total and incident CRC cases in the total invited group were reduced accordingly (Table 1, lower part).

Calculation of relative risks

We derived estimates of relative risks of both total (prevalent and incident) CRC and of incident CRC according to both intention-to-screen and per-protocol analysis. Intention-to-screen estimates, which quantify the impact of the screening offer on the CRC risk, were derived as ratios of case numbers with and without screening in the invited group. Adjusted per-protocol estimates, which quantify the impact of actual use of screening on the CRC risk, were derived as ratios of case numbers with and without screening in the screened group.

This approach of deriving adjusted per-protocol estimates is unaffected by potential differences in baseline risk between screening attenders and non-attenders and corresponds to such adjustment according to the method proposed by Cuzick et al. [13] used in sensitivity analyses of the original NordICC trial report. We derived 95% confidence intervals for our relative risk estimates as the 2.5^th and 97.5^th percentile of one million runs of Monte Carlo simulations of the NordICC trial, using the observed case proportions as expected values for each simulation run. A more general formal description of the derivation of relative risks is provided in the Supplementary Material 1, and the R code used for the Monte Carlo simulations is provided in the Supplementary Material 2.

Besides enabling separate estimates of screening effects for incident (potentially preventable) CRC cases, our approach slightly differs from the original analysis of the NordICC trial in that relative risk estimates were based on count data rather than incidence rate (person-time) data. In order to judge the relevance of this slight difference, we present the relative risk estimates for total CRC (prevalent and incident CRC combined) from the original publication side-by-side with our estimates.

Table 3 shows the effect estimates with 95% confidence intervals for total (prevalent and incident) CRC that were reported in the original trial publication [4] and recalculated effect estimates, separately for total CRC and incident (potentially preventable) CRC in intention-to-screen and per-protocol analysis. Our relative risk estimates for the offer of screening colonoscopy in intention-to-screen analysis and for the use of screening colonoscopy in adjusted per-protocol analysis for total (prevalent and incident) CRC were 0.83 (95% CI 0.72-0.96) and 0.66 (95% CI 0.50-0.91), and they closely match corresponding estimates reported in the original NordICC trial publication [4]. However, much stronger effects were estimated in analyses focusing on truly incident, potentially preventable CRC cases, with relative risk estimates of 0.72 (95% CI 0.59-0.91) in intention-to-screen analysis and 0.43 (0.26-0.78) in adjusted per-protocol analysis.

Variation of the share of prevalent cases among all CRC cases among non-attenders of screening in the invited group (between 30% and 50% compared to 40% among attenders in the absence of screening effects), while maintaining Assumption 1, led to minor variation of the intention-to-screen relative risk estimate for truly incident CRC between 0.74 and 0.69 and did not lead to any changes in the adjusted per-protocol effect estimates.

In this article, we demonstrate that the effects of screening colonoscopy in preventing incident CRC are much stronger than suggested by reported analyses from the NordICC trial, the so far only randomized trial reporting on reduction of CRC incidence and mortality by the offer of screening colonoscopy. We show that reported effect estimates of CRC incidence reduction have been strongly diluted by inclusion of prevalent, no longer preventable CRC cases. Exclusion of prevalent cases from the outcome measures in both the invited group and the usual-care group increased estimated risk reduction by the offer of screening colonoscopy from 17% to 28% in intention-to-screen analysis. Estimated risk reduction by the use of screening colonoscopy in adjusted per-protocol analysis even increased from 34% to 57%.

That „prevalence bias“ most likely led to major underestimation of the reduction of CRC incidence in the NordICC trial has been pointed out in preliminary work that had relied on consideration of additional external data, including a modelling study [9-12]. In a previous preliminary analysis, we had derived a theoretically possible range and a plausible range of the reduction of truly incident CRC from published results of the trial and national cancer registry data [9]. Although effect estimate ranges for screening offer and screening use suggested stronger effects than those reported from the original NordICC study publication, the ranges were rather wide: 0.14 to 0.62 and 0.41 to 0.59 for the theoretically possible and the plausible range of the point estimates of relative risk in per-protocol analyses, and no confidence intervals for the effect estimates were available. Our now derived point estimates, which are exclusively based on data reported from the NordICC trial, are well within the previously derived plausible range and are provided with 95% confidence intervals, thereby accounting for statistical uncertainty of the effect estimates for the reduction of truly incident cases. Our now-derived relative risk estimate in the intention-to-screen analysis (0.72) also almost perfectly matches the estimate derived from a recent modelling study [12]. Although that modelling study had suggested an even stronger screening effect in the per-protocol analysis, the relative risk estimate of 0.30 in the modelling study was well within the 95% confidence interval of the effect derived in our current analysis which is independent of additional assumptions that had to be made in the modelling study.

The main reason why prevalent CRC cases had been included in the incidence-based outcome measures in the design of the NordICC trial and the reporting of its results is that such prevalent cases are directly observable and their number is known with certainty for attenders of screening colonoscopy only [14]. In fact, for this subgroup of the trial population, prevalent cases even made up the majority, i.e. 62 out of 102 (61%) of reported „incident cases“. In this article, we demonstrate how proportions and numbers of prevalent CRC cases can be estimated under plausible assumptions also among non-attenders of screening colonoscopy, in whom they are not directly observable. Although these proportions were lower than among screened participants, they still made up approximately 40% of CRC cases.

Our analyses are based on two assumptions which deserve careful discussion. The first assumption was that proportions of participants with prevalent and incident CRC cases would have been equal in the invited group and the usual-care group in the absence of screening effects. This assumption seems highly plausible given the randomized study design and the large number of study participants. Of course, we cannot rule out some random variation which would also have affected the results in the original study report.

The second assumption was that, in the absence of screening, the share of prevalent CRC cases among all CRC cases would have been the same in attenders and non-attenders of screening. It is worth noting that this assumption does not require that absolute prevalences and incidences would have been the same in attenders and non-attenders. In the contrary, our analyses explicitly took care of potentially different uptake of screening according to the participants' CRC risk and assumed that, in the absence of screening, a higher baseline risk in either group would go along with both higher prevalence and incidence of CRC in that group, which seems plausible. Furthermore, as shown in sensitivity analyses, even major violation of this assumption would lead to only modest variation of the intention-to-screen estimates of relative risk of truly incident CRC, and it would not affect the adjusted per-protocol estimates. The latter could therefore be estimated by just relying on the “standard RCT assumption”, i.e. equal risks among the invited and the usual-care group due to randomization.

Our analysis focused on the so far only RCT reporting effects of screening colonoscopy on CRC incidence. The results reported in late 2022 reflect interim results after 10 years of follow-up, and final analyses are planned after 15 years of follow-up. With increasing length of follow-up, the share of prevalent cases among all CRC cases will decrease, and the underestimation of effects on truly incident cases is expected to become smaller. Nevertheless, some relevant underestimation will still be there in planned conventional analyses, and we encourage to employ the type of analyses proposed in this article to more fully disclose the impact of screening colonoscopy in preventing incident CRC in the final analysis. It is also worth noting that underestimation of true screening effects on CRC incidence is not unique to the NordICC trial but also has affected previous RCTs on CRC screening. In particular, all of the four large RCTs on CRC screening by flexible sigmoidoscopy have likewise included prevalent CRC in their outcome measures of CRC incidence [15-19], and thereby most likely also substantially underestimated the preventive effects of screening [20].

Our analysis focused on CRC incidence, one of the two primary outcomes of the NordICC trial. Obviously, prevalence bias would not affect effect estimates for CRC mortality, the other primary outcome measure. In the contrary, detection of prevalent cases would be expected to contribute to CRC mortality reduction, in addition to such reduction by prevention of incident cases. Early detection of prevalent cases would even be expected to be the main contributor to mortality reduction in the early years of follow-up.

However, effect estimates on both CRC incidence and mortality in the NordICC trial and the flexible sigmoidoscopy trials might have further been diluted by other factors, such as use of diagnostic colonoscopies outside the screening offer [21,22]. According to representative population surveys, substantial proportions of the older population in European countries, including the countries involved in the NordICC trial, meanwhile have had at least one colonoscopy within 10 years, the follow-up period of the NordICC trial report [23,24]. Such colonoscopies would be expected to have similar protective effects as the screening colonoscopies offered within the trial, as precursors of CRC, such as adenomas or sessile serrated polyps, would typically be also removed at such diagnostic colonoscopies. This may have further substantially aggravated underestimation of screening effects [21]. Along with inclusion of non-preventable (prevalent) cancers in the effect estimates, such additional issues may explain the discrepancy between the apparently very modest effect estimates reported in the NordICC trial and the „real life“ strong decline of CRC incidence in the screening age populations in the US and some European countries, in which prevalence of screening colonoscopy has strongly increased in recent years [8,25].

In conclusion, our analyses suggest that prevention of incident CRC by detection and removal or precancerous lesions at screening colonoscopy is much stronger than suggested by reported results of the NordICC trial, the so far only RCT on long-term effects of screening colonoscopy. The reported lower-than-expected reduction of CRC incidence in this trial should not unduly discourage screening efforts to curb the CRC epidemic that is otherwise to be expected from the demographic changes [26,27] and unfavorable trends in the prevalence of some key risk factors, such as overweight and obesity [28]. Effective reduction of CRC incidence and mortality may be achieved by various screening approaches, such as screening by fecal immunochemical test, flexible sigmoidoscopy or colonoscopy. It has already been achieved to a remarkable extent in some countries in which either of these methods is widely used [8,25,29], whereas CRC incidence has kept rising in many other countries [25,30]. We hope that our demonstration of stronger effects than those suggested by the original NordICC trial publication will help to encourage wider implementation and use of effective CRC screening offers.

Statements and Declarations

Funding. This work was supported in part by grants from the German Federal Ministry of Education and Research (grant no. 01KD2104A) and the German Cancer Aid (grant no. 70114735).

Competing Interests. The authors have no relevant financial or non-financial interests to disclose.

Author Contributions.HB designed the study. HB and TH-L conducted the analyses. HB drafted the manuscript, MH and HB drafted Figure 1. All authors reviewed the draft, had access to and verified the data, contributed to the revision of the manuscript and agreed to its submission for publication.

Ethics Approval. Not applicable. This study included only calculations based on previously published summary statistics of a multicenter randomized controlled trial that had obtained ethical approval by the ethics committees as reported in the original publication from that trial.

Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med 1993;329(27):1977-81.
Brenner H, Stock C, Hoffmeister M. Effect of screening sigmoidoscopy and screening colonoscopy on colorectal cancer incidence and mortality: systematic review and meta-analysis of randomised controlled trials and observational studies. BMJ 2014;348:g2467.
Lauby-Secretan B, Vilahur N, Bianchini F, Guha N, Straif K; International Agency for Research on Cancer Handbook Working Group. The IARC Perspective on Colorectal Cancer Screening. N Engl J Med 2018;378(18):1734-40.
Bretthauer M, Løberg M, Wieszczy P, et al. Effect of Colonoscopy Screening on Risks of Colorectal Cancer and Related Death. N Engl J Med 2022;387(17):1547-56.
The Lancet Gastroenterology Hepatology. Controversy over colonoscopy for colorectal cancer screening. Lancet Gastroenterol Hepatol 2022;7(12):1061.
Powell K, Prasad V. Interpreting the results from the first randomised controlled trial of colonoscopy: does it save lives? BMJ Evid Based Med 2023;28(5):306-8.
Ran T, Cheng CY, Misselwitz B, Brenner H, Ubels J, Schlander M. Cost-Effectiveness of Colorectal Cancer Screening Strategies - A Systematic Review. Clin Gastroenterol Hepatol 2019;17(10):1969-81.e15.
Siegel RL, Miller KD, Goding Sauer A, et al. Colorectal cancer statistics, 2020. CA Cancer J Clin 2020;70(3):145-64.
Brenner H, Heisser T, Cardoso R, Hoffmeister M. When gold standards are not so golden: prevalence bias in randomized trials on endoscopic colorectal cancer screening. Eur J Epidemiol 2023;38(9):933-7.
Brenner H, Hoffmeister M. Re: Interpreting epidemiologic studies of colorectal cancer prevention. Eur J Epidemiol 2024;39(1):97.
Brenner H, Heisser T, Cardoso R, Hoffmeister M. Reduction in colorectal cancer incidence by screening endoscopy. Nat Rev Gastroenterol Hepatol 2024;21(2):125-133.
Heisser T, Hoffmeister M, Brenner H. Significant underestimation of preventive effects in colorectal cancer screening trial. Gut 2024;73(3):556-558.
Cuzick J, Edwards R, Segnan N. Adjusting for non-compliance and contamination in randomized clinical trials. Stat Med 1997;16(9):1017-29.
Song M, Bretthauer M. Authors' Reply: Interpreting epidemiologic studies of colorectal cancer prevention. Eur J Epidemiol 2024;39(1):99.
Atkin W, Wooldrage K, Parkin DM, et al. Long term effects of once-only flexible sigmoidoscopy screening after 17 years of follow-up: the UK Flexible Sigmoidoscopy Screening randomised controlled trial. Lancet 2017;389(10076):1299-311.
Holme Ø, Schoen RE, Senore C, et al. Effectiveness of flexible sigmoidoscopy screening in men and women and different age groups: pooled analysis of randomised trials. BMJ 2017;356:i6673.
Miller EA, Pinsky PF, Schoen RE, Prorok PC, Church TR. Effect of flexible sigmoidoscopy screening on colorectal cancer incidence and mortality: long-term follow-up of the randomised US PLCO cancer screening trial. Lancet Gastroenterol Hepatol 2019;4(2):101-10.
Senore C, Riggi E, Armaroli P, et al. Long-Term Follow-up of the Italian Flexible Sigmoidoscopy Screening Trial. Ann Intern Med 2022;175(1):36-45.
Juul FE, Cross AJ, Schoen RE, et al. 15-Year Benefits of Sigmoidoscopy Screening on Colorectal Cancer Incidence and Mortality: A Pooled Analysis of Randomized Trials. Ann Intern Med 2022;175(11):1525-33.
Brenner H, Heisser T, Cardoso R, Hoffmeister M. 15-Year Benefits of Sigmoidoscopy Screening on Colorectal Cancer Incidence and Mortality. Ann Intern Med 2023;176(6):eL230024.
Brenner H, Stock C, Hoffmeister M. In the era of widespread endoscopy use, randomized trials may strongly underestimate the effects of colorectal cancer screening. J Clin Epidemiol 2013;66(10):1144-50.
Hoffmeister M, Cardoso R, Brenner H. Colonoscopy Screening and Colorectal Cancer Incidence and Mortality. N Engl J Med 2023;388(4):377-8.
Cardoso R, Guo F, Heisser T, Hoffmeister M, Brenner H. Utilisation of Colorectal Cancer Screening Tests in European Countries by Type of Screening Offer: Results from the European Health Interview Survey. Cancers 2020;12(6):1409.
Eurostat. Self-reported last colonoscopy by sex, age and educational level. https://ec.europa.eu/eurostat/databrowser/view/HLTH_EHIS_PA6E__custom_3573783/default/table?lang=en), accessed on August 11, 2023.
Cardoso R, Guo F, Heisser T, et al. Colorectal cancer incidence, mortality, and stage distribution in European countries in the colorectal cancer screening era: an international population-based study. Lancet Oncol 2021;22(7):1002-13.
Heisser T, Hoffmeister M, Tillmanns H, Brenner H. Impact of demographic changes and screening colonoscopy on long-term projection of incident colorectal cancer cases in Germany: A modelling study. Lancet Reg Health Eur 2022;20:100451.
Morgan E, Arnold M, Gini A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN. Gut 2023;72(2):338-44.
Jaacks LM, Vandevijvere S, Pan A, et al. The obesity transition: stages of the global epidemic. Lancet Diabetes Endocrinol 2019;7(3):231-40.
Australian Institute of Health and Welfare. National Bowel Cancer Screening Program monitoring report 2022, catalogue number CAN 148, AIHW, Australian Government (2022). https://www.aihw.gov.au/reports/cancer-screening/nbcsp-monitoring-2022/summary. Last accessed February 29, 2024.
GBD 2019 Colorectal Cancer Collaborators. Global, regional, and national burden of colorectal cancer and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Gastroenterol Hepatol 2022;7(7):627-47.

Table 1. Numbers of CRC cases reported by Bretthauer et al [4] and expected numbers of prevalent and incident CRC cases without and with screening effect^a

Data	Group		Participants	CRC cases
Data	Group		Participants	Total	Prevalent	Incident

Reported [4]	Usual-care	Total	56,365	622
	Invited	Total	28,220	259
		Non-attenders	16,377	157
		Attenders	11,843	102	62	40

Expected if no screening effect^a	Usual-care	Total	56,365	622	250	372
	Invited	Total	28,220	311	125	186
		Non-attenders	16,377	157	63	94
		Attenders	11,843	154	62	92

Expected with screening effect^a	Usual-care	Total	56,365	622	250	372
	Invited	Total	28,220	259	125	134
		Non-attenders	16,377	157	63	94
		Attenders	11,843	102	62	40

CRC, colorectal cancer

^a Expected under the assumptions listed in Table 2.

Table 2. Assumptions made for calculating screening effects on reduction of truly incident cases

Assumption 1

In the absence of screening effects, the proportions of participants with prevalent and incident CRC cases would have been equal in the invited group and the usual-care group.

Assumption 2

(needed for intention-to-screen estimates of reduction of truly incident cases only)

In the absence of screening effects, the share of prevalent CRC cases among all CRC cases would have been the same among screening non-attenders as among screening attenders in the invited group.

Sensitivity analyses:

10 percent units higher or lower share of prevalent cases among non-attenders

CRC, colorectal cancer

Table 3. Reported effect estimates for total (prevalent and incident) CRC reported by Bretthauer et al [4] and recalculated effect estimates of total CRC and incident CRC in intention-to-screen and per-protocol analysis.

CRC cases	Analysis	Relative risk (95% confidence interval)

Total (prevalent and incident)	Intention-to-screen, reported [4]	0.82 (0.70-0.93)
	Intention-to-screen, recalculated	0.83 (0.72-0.96)

	Per-protocol, reported⁴	0.66 (0.46-0.86)^a
	Per-protocol, recalculated	0.66 (0.49-0.90)

Incident	Intention-to-screen, main analysis^b	0.72 (0.59-0.91)
	Intention-to-screen, sensitivity analysis 1^c	0.74 (0.61-0.94)
	Intention-to-screen, sensitivity analysis 2^d	0.69 (0.57-0.88)

	Per-protocol,, main analysis^b	0.43 (0.26-0.78)
	Per-protocol, sensitivity analysis 1^c	0.43 (0.26-0.78)
	Per-protocol, sensitivity analysis 2^d	0.43 (0.26-0.78)

CRC, colorectal cancer

^a adjusted by the method proposed by Cuzick et al. [13]

^bassuming equal share (40%) of prevalent CRC cases among all CRC cases among non-attenders and attenders of screening in the absence of screening effects

^cassuming lower share of prevalent CRC cases among all CRC cases among non-attenders (30%) than among attenders (40%) of screening in the absence of screening effects

^dassuming higher share of prevalent CRC cases among all CRC cases among non-attenders (50%) than among attenders (40%) of screening in the absence of screening effects

Effects of screening colonoscopy in preventing colorectal cancer:
re-analysis of the NordICC trial

Status:

Version 1

Abstract

Introduction

Materials And Methods

Results

Discussion

Declarations

References

Tables

Supplementary Files

Status:

Version 1

Effects of screening colonoscopy in preventing colorectal cancer:re-analysis of the NordICC trial

Status:

Version 1

Abstract

Introduction

Materials And Methods

Results

Discussion

Declarations

References

Tables

Supplementary Files

Status:

Version 1

Effects of screening colonoscopy in preventing colorectal cancer:
re-analysis of the NordICC trial